Agent Based Models: Introduction

Ernesto Carrella

April 18, 2019

Agent Based Models

  • Define
    • Individuals
    • Interactions
  • Observe Emergent Patterns

Traffic Example

Traffic Example - 2

When should we bother?

  • Benefits:
    • Knowledge is only available at individual level
    • When aggregation is too hard to model mathematically
    • When we want to test what would generate an aggregate pattern
  • Cost:
    • Intractable likelihood

Shelling model 1

Shelling model 2

Individual Level Policy - 1

Individual Level Policy - 2

Squaring with data

  • Estimation problems
    • Intractable likelihood
    • Many parameters
  • Techniques
    • Approximate Bayesian Computation
    • Indirect Inference
    • Synthetic Likelihood

Squaring with data 2

Squaring with data 3

Microsimulation

  • Gateway drug to ABM
  • EUROMOD at Exeter
  • Data focused:
    • heterogeneity
    • no interactions

Microsimulation - 1

Microsimulation - 2

Microsimulation - 3

Suggested Reading - 1

Suggested Reading - 2

Suggested Reading - 3